1. Technical Field
The present disclosure relates to a technology of correcting two-dimensional images, and more particularly to a method for correcting vignetting effects of a two-dimensional image by estimating a vignetting function representing the vignetting effects of the two-dimensional image and an apparatus for the same.
2. Related Art
In imaging technology domains, a ‘vignetting’ usually means an effect in which intensity of lights is attenuated radially from an optical center of an image. Such the vignetting effects may prevent scenes included in the image from being correctly delivered to viewers.
Therefore, various methods for correcting vignetting effects of an image are being studied in the imaging technology domains. The most widely-used vignetting correction method is a method based on photometric calibration which measures a unique vignetting function of a camera by using a template image predefined according to setting and lenses of the camera. However, the method based on the photometric calibration can remove vignetting effects only corresponding to the setting and lenses of the camera, and they cannot perform vignetting correction on arbitrary images.
On the other hand, vignetting correction methods using a plurality of images having different intensity attenuation are being used. Since such the methods use a property that different images captured by using the same camera have the same vignetting function, the predefined template images are not necessary. However, since the methods need a plurality of images, there is a problem that vignetting correction on an arbitrary single image downloaded from an internet is difficult.
Recently, vignetting correction methods for an arbitrary single image have been introduced. For example, a vignetting function estimation method using intensity attenuation of respective areas segmented by performing image segmentation techniques on areas having homogeneous color or texture has been introduced. Also, a vignetting function estimation method using symmetry of a radial gradient distribution of an image has been introduced.
Although the above vignetting estimation and correction methods using a single image can be flexibly applied to various images, the method based on image segmentation is weak to noise of the image, and the method using symmetry of a radial gradient distribution of an image has high computational complexity and is sensitive to numerical errors. Especially, since such the methods use two-dimensional image optimization, it takes too much time for the methods to perform vignetting function estimation.